Cohere CEO Gomez Calls Rival Labs' Standards Plan a 'Cartel'

Cohere co-founder and CEO Aidan Gomez, in a September 13 essay on his company's site, accuses a group of 'market-dominant AI companies from Silicon Valley' of pushing an industry-run standards body that would amount to 'a cartel by any other name.'

Gomez names Anthropic CEO Dario Amodei and frames the broader target as 'two or three Silicon Valley companies' asking governments for 'a narrow antitrust waiver to make that coordination lawful.' The result, he writes, would leave 'every other AI developer to blindly follow whatever the participants settle on.' His essay follows the AI Weekly tracker's report earlier this week that Anthropic, OpenAI and Google have been meeting since July on an AI standards body.

His central objection is that self-written rules ossify the incumbents' current lead. 'A safety regime designed by a few labs will only be rigorous about the risks they have already built their safety systems to assess,' he writes, and it 'turns their current advantages into baseline for what it takes to compete safely.' He also takes aim at the credibility of the safety pitch itself: 'The failures we saw reported in July happened inside the two best-resourced labs in the world.'

Gomez then reaches for two historical parallels. In 1975 the SEC designated 'three firms as Nationally Recognized Statistical Rating Organizations and never published criteria for how anyone else might earn the designation'; the same firms, he writes, later 'rated subprime mortgage securities triple-A and nearly took the global economy down.' In 1985 European carmakers won 'a sweeping antitrust waiver' under the Motor Vehicle Block Exemption on safety grounds, and 'it took the European Commission roughly twenty five years of reforms to unwind' it.

In place of a closed body, Gomez proposes four pillars: an evidence-based, internationally published risk framework funded through public research bodies; mandatory developer transparency via model cards and incident reporting; independent testing scoped only to 'genuinely dangerous capabilities' such as cyberattacks, fraud, bioweapons and critical infrastructure; and independent assurance mechanisms modeled on financial, aviation and nuclear regulation. 'In both cases, the stated goal was safety,' he writes of the bond and auto analogies. 'But the result was a market structure that protected incumbents and limited competition.'